- So we're creating truth table from this one.
- For example if x1 and x2 is either 1 or 0, and both weights are 1/2. Suppose one equals zero, what are the value of the others to be at least equals the threshold?
- We con do some equation, either x1 or x2 are zero. and create the point.
- The result will be a plane that drawn to separate the y=0/y=1 threshold

- We have specified the line separator for the plane. And if we plot x1,x2 like truth table. We ended up to must have both of them equals 1 to be in the green zone. And that's makes an "AND" table

- Perceptron training is training our perceptron. It will be wastefull to just do perceptron by hand.
- There's two options, perceptron rule or gradient descent.

- Now what we do is some kind of other supervised learning, where we structure our data to have x data set and coressponding parameters, with the output as our data target.
- Then we perform iteration in curly braces. Each weight will be increasing every iterations.
- The delta weight will be learning rate . (y-h(x) . x data set
- here, the H(x) is the activation unit, and the threshold got subtittute to the left side as bias unit.
- If we cand find linear separable from the data, then the neural networks will find it.
- Otherwise it goes to infinite loop.